Deep learning prediction of stroke thrombus red blood cell content from multiparametric MRI

BACKGROUND AND PURPOSE

Thrombus red blood cell (RBC) content has been shown to be a significant factor influencing the efficacy of acute ischemic stroke treatment. In this study, our objective was to evaluate the ability of convolutional neural networks (CNNs) to predict ischemic stroke thrombus RBC content using multiparametric MR images.

MATERIALS AND METHODS

Retrieved stroke thrombi were scanned ex vivo using a three-dimensional multi-echo gradient echo sequence and histologically analyzed. 188 thrombus R2*, quantitative susceptibility mapping and late-echo GRE magnitude image slices were used to train and test a 3-layer CNN through cross-validation. Data augmentation techniques involving input equalization and random image transformation were employed to improve network performance. The network was assessed for its ability to quantitatively predict RBC content and to classify thrombi into RBC-rich and RBC-poor groups.

RESULTS

The CNN predicted thrombus RBC content with an accuracy of 62% (95% CI 48-76%) when trained on the original dataset and improved to 72% (95% CI 60-84%) on the augmented dataset. The network classified thrombi as RBC-rich or poor with an accuracy of 71% (95% CI 58-84%) and an area under the curve of 0.72 (95% CI 0.57-0.87) when trained on the original dataset and improved to 80% (95% CI 69-91%) and 0.84 (95% CI 0.73-0.95), respectively, on the augmented dataset.

CONCLUSIONS

The CNN was able to accurately predict thrombus RBC content using multiparametric MR images, and could provide a means to guide treatment strategy in acute ischemic stroke.

Multimodality Characterization of the Clot in Acute Stroke

Aim: Current treatment of occluded cerebral vessels can be done by a variety of endovascular techniques. Sometimes, the clot responds in varying degrees to the treatment chosen. The Ex vivo characterization of the clot occluding the arteries in acute ischemic stroke can help in understanding the underlying imaging features obtained from pre-treatment brain scans. For this reason, we explored the potential of microCT when combined with electron microscopy for clot characterization. Results were compared to the clinical CT findings.

Methods: 16 patients (9 males, 8 females, age range 54–93 years) who were referred to our institution for acute stroke underwent dual-source CT.

Results: Clinical CT clots were seen as either iso or hyperdense. This was corroborated with micro-CT, and electron microscopy can show the detailed composition.

Conclusion: MicroCT values can be used as an indicator for red blood cells-rich composition of clots. Meaningful information regarding the clot composition and modalities of embedding along the stent retrievers can be obtained through a combination of microCT and electron microscopy.

Correlation between acute ischaemic stroke clot length before mechanical thrombectomy and extracted clot area: Impact of thrombus size on number of passes for clot removal and final recanalization

Introduction

We assessed the correlation between thrombus size before and after mechanical thrombectomy, measured as length by Computed Tomography Angiography/Non-Contrast Computed Tomography (CTA/NCCT) and Extracted Clot Area, ECA, respectively. We also assessed the influence of thrombus size on the number of passes required for clot removal and final recanalization outcome.

Materials and methods

Acute ischaemic stroke (AIS) thrombi retrieved by mechanical thrombectomy from 500 patients and data of clot length by CTA/NCCT were collected from three hospitals in Europe. ECA was obtained by measuring the area of the extracted clot. Non-parametric tests were used for data analysis.

Results

A strong positive correlation was found between clot length on CTA/NCCT and ECA (rho = 0.619,N = 500,P < 0.0001*). Vessel size influences clot length on CTA/NCCT (H2 = 98.6, P < 0.0001*) and ECA (H2 = 105.6,P < 0.0001*), but the significant correlation between CTA/NCCT length and ECA was evident in all vessels. Poorer revascularisation outcome was associated with more passes (H5 = 73.1, P < 0.0001*). More passes were required to remove longer clots (CTA/NCCT; H4 = 31.4, P < 0.0001*; ECA; H4 = 50.2, P < 0.0001*). There was no significant main association between recanalization outcome and length on CTA/NCCT or ECA, but medium sized clots (ECA 20–40 mm²) were associated with least passes and highest revascularisation outcome (N = 500, X² = 16.2, P < 0.0001*).

Conclusion

Clot length on CTA/NCCT strongly correlates with ECA. Occlusion location influences clot size. More passes are associated with poorer revascularisation outcome and bigger clots. The relationship between size and revascularisation outcome is more complex. Clots of medium ECA take less passes to remove and are associated with better recanalization outcome than both smaller and larger clots.

Hyperdense vessel sign as a potential guide for the choice of stent retriever versus contact aspiration as first-line thrombectomy strategy

BACKGROUND

The first-pass effect (FPE) has emerged as a key metric for efficacy in mechanical thrombectomy (MT). The hyperdense vessel sign (HDVS) on non-contrast head CT (NCCT) indicates a higher clot content of red blood cells.

OBJECTIVE

To assess whether the HDVS could serve as an imaging biomarker for guiding first-line device selection in MT.

METHODS

A prospective MT database was reviewed for consecutive patients with anterior circulation large vessel occlusion stroke who underwent thrombectomy with stent retriever (SR) or contact aspiration (CA) as first-line therapy between January 2012 and November 2018. Pretreatment NCCT scans were evaluated for the presence of HDVS. The primary outcome was FPE (modified Thrombolysis in Cerebral Infarction score 2c/3). The primary analysis was the interaction between HDVS and thrombectomy modality on FPE. Secondary analyses aimed to evaluate the predictors of FPE.

RESULTS

A total of 779 patients qualified for the analysis. HDVS and FPE were reported in 473 (60.7%) and 286 (36.7%) patients, respectively. The presence of HDVS significantly modified the effect of thrombectomy modality on FPE (p=0.01), with patients with HDVS having a significantly higher rate of FPE with a SR (41.3% vs 22.2%, p=0.001; adjusted OR 2.11 (95% CI 1.20 to 3.70), p=0.009) and non-HDVS patients having a numerically better response to CA (41.4% vs 33.9%, p=0.28; adjusted OR 0.58 (95% CI 0.311 to 1.084), p=0.088). Age (OR 1.01 (95% CI 1.00 to 1.02), p=0.04) and balloon guide catheter (OR 2.08 (95% CI 1.24 to 3.47), p=0.005) were independent predictors of FPE in the overall population.

CONCLUSION

Our data suggest that patients with HDVS may have a better response to SRs than CA for the FPE. Larger confirmatory prospective studies are warranted.

The role of advanced neuroimaging techniques in ischemic stroke prevention

In great part due to recent advances in medical technology, ischemic cerebral stroke has become an increasingly treatable condition, which requires urgent measures and which rely on pharmacological and/or interventional measures. Due to its high prevalence, preventive measures should also be undertaken, and this is a situation where the use of advanced neuroimaging techniques could be helpful in certain underlying diseases. What is proposed here is to discuss how modern neuroimaging techniques (advanced magnetic resonance (MR) techniques and/or nuclear medicine techniques such as positron emission tomography (PET)) could help in situations that would otherwise lead to a stroke. Since both primary and secondary prevention measures are often required, we see that the techniques can be helpful in both situations. The diseases that cause stroke that can be investigated are, among others, carotid stenosis; transient ischemic attacks (TIAs) may also be followed by a major stroke if nothing is undertaken. It has been established that carotid stenosis is to be treated in the presence of a significant lesion that has neurological symptoms. The question of how to deal with these patients often arises when the relationship between the stenosis and symptoms is not significant or the symptomatology unclear. In such situations, either PET and/or fat-saturated T1 images of the carotids can help to demonstrate the embolic nature of the plaque. We have seen that carotid plaque vulnerability, which can cause embolism, can be associated with plaque inflammation (seen on PET) or plaque haemorrhage (seen on MR images). Also, while PET and MRI will demonstrate different stages of plaque vulnerability, they can both help to demonstrate vascular lesions that are at risk of causing significant ischemic events. Diffusion-weighted imaging (DWI) has shown that some TIAs may indeed be ischemic brain lesions with a transient symptomatology. The early use of DWI can thus help treat these patients more acutely. Based on this, we have seen that newer imaging techniques can provide additional knowledge about conditions that may lead to stroke and be treated. This should have a major impact on patient outcomes and ultimately on healthcare costs related to this condition.